GITpy: A Python Implementation of the Generalized Inversion Technique

GITpy is an open-source object-oriented Python software package implementing the well-established Generalized Inversion Technique (GIT), a spectral decomposition approach to isolate the source, propagation, and site contributions from S-phase Fourier amplitude spectra (FAS) (Andrews, 1986; Castro et al., 1990; Boatwright et al., 1991; Drouet et al., 2008; Edwards et al., 2008; Oth et al., 2011; Bindi et al., 2020). GITpy applies a nonparametric (i.e., without imposing any a priori parametric models on the different terms), one-step inversion procedure. GITpy offers the possibility to: (1) simplify the attenuation and source modeling process by providing configuration files and interactive procedures allowing for rapid testing of different models; (2) choose between different levels of attenuation modeling complexity (geometrical spreading and anelastic attenuation); (3) select among different source spectrum modeling options, including the use of a homogeneous or heterogeneous crustal model, as well as the ability to define the frequency range for the model fitting; (4) calculate station-specific apparent source spectra by correcting the input FAS for site amplification and nonparametric attenuation obtained from the inversion, and then fit them. This can be particularly useful for directivity studies. Furthermore, this module can be used independently for a rapid estimation of source parameters in case of a strong event; (5) provide several source parameters including radiated energy, apparent stress, and radiation efficiency alongside seismic moment, corner frequency, and stress drop. Here, we present the versatility of GITpy by applying it to the well-documented 2016–2017 seismic sequence in central Italy, showcasing the software’s capabilities through specific modules for source and attenuation modeling, as well as for calculating apparent source spectra. To achieve this, a comprehensive dataset including 355 stations and 8534 events was assembled, allowing for the evaluation of the software’s performance in handling large-scale datasets.